On Sat, 26 Nov 2011, Gabor Ruzsa wrote:
Dear Gretl users,
I'm totally puzzled by the option in Gretl to include lags of the dependent variable
in
time series models estimated via FGLS (Cochrane-Orcutt or Prais-Winsten). As much as I
know, the consistence of the FGLS estimator requires the strict exogeneity of the
regressors, i.e. including lagged dependent variables is out of question.
Although I can't think of any reference right now, nor improvise a proof
of inconsistency, my gut feeling is that you're absolutely right. However,
restricting a software package to do only the "right" things would be a
very bad idea. First, because what it considered "right" changes in time
(and besides, who says what's right and what's wrong? --- but I digress).
Second, because sometimes you _want_ to do wrong things on purpose, either
for pedagogical purposes or just for fun. Third, because in certain cases
you want to replicate other people's results, even when they do wrong
things.
So, gretl leaves you the option of being naughty or nice.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti